Delete vaiables in PCA analysis


I have been approaching PCA analysis for a few months and I have a question to ask.
We are discussing with a colleague of mine whether or not it is right to eliminate variables from our model characterized by spectroscopic data.

Plotting the first 2 principal components, these bring in our case 27.8% of the total variance of the system.
Is it a correct approach to eliminate the "Loadings" which have a minor contribution in the description of these 2 components?

In this way, the R-square of the 2 components clearly increases, but have we not arbitrarily decided in this way that in the description of our phenomenon under examination these 2 components have a greater incidence than in the real conditions? We are choosing to eliminate variables that carry information for all the other components, clearly forcing the system to give greater weight to the first and second main components.

Translated to the case in question, we have increased the incidence of the first component (and of related phenomen, in our case a biological treatment) from 15.7% to 39.8% on the whole system

Spectroscopic datas suffer from noise, but is this a correct approach to data cleaning (to increase R-2)?

I hope I got it right,
Thank you!